Paul Price added a comment - 03/Aug/16 10:59 AM Scott Daniel , Simon said you're using GalSim on the sims side. Please let me know if this will trouble you. I added a patch on tickets/ DM-7114 of our galsim product, and that's working its way through Jenkins. The GalSim PR I submitted might be changed to fix the condition that's causing the assertion to fail. I don't think we care which way it goes since we've got our own fix, but we'll want to remove this fix the next time we upgrade galsim.

Scott Daniel added a comment - 03/Aug/16 11:09 AM Paul Price Thanks for the heads-up. That should be fine. As long as all of the sims_GalSimInterface unit tests pass (which, I suspect, they will), this won't affect us.

Paul Price added a comment - 03/Aug/16 11:12 AM The Travis tests for the GalSim PR passed, so it should be fine.

Scott Daniel added a comment - 03/Aug/16 11:14 AM I was referring to the unit tests in one of our sims packages. Regardless, I am quite certain this won't affect us.

Paul Price added a comment - 03/Aug/16 11:26 AM Oh, I see; sorry. I see that sims_GalSimInterface passed in Jenkins : sims_GalSimInterface: 12.0-2-gb27458b+2 ......................................................................................................ok (239.4 sec).

Paul Price added a comment - 03/Aug/16 11:31 AM John Parejko , would you mind reviewing this? It's quite small (14 lines), though there's a catch: the patch is itself a patch, so it's a little meta. price@price-laptop:~/LSST/external/galsim (tickets/DM-7114=) $ git sub commit 24fbce360ac8c9ccc7c4719e18cd964fb12ff2fc Author: Paul Price <price@astro.princeton.edu> Date: Wed Aug 3 12:36:24 2016 -0400   replace assertion with exception This is an exceptional condition that can occur sometimes (because it certainly does occur!), rather than something that should bring the world to a screaming halt. Reported upstream as PR #784: https://github.com/GalSim-developers/GalSim/pull/784   patches/DM-7114-assertion_to_exception.patch | 14 ++++++++++++++ 1 file changed, 14 insertions(+)

John Parejko added a comment - 03/Aug/16 11:56 AM Seems fine, though I'd like the exception to say something about what bounds (iy1, iy2 or what those mean higher up) and/or include the actual values. I see the debug statements above the throw, but it's nice for the exception message to be self contained.

John Parejko added a comment - 03/Aug/16 11:58 AM Oh, wait. The assert is still there at the bottom: shouldn't the assert(iy1 <= iy2); line be removed, since it's already been handled by the except?

Paul Price added a comment - 03/Aug/16 12:30 PM Thanks John Parejko . I would add more information to the exception, except that that's not done for any of the other exceptions in that file, and so I'm not sure what technique they would prefer for string formatting and I don't want to add any #include files just for this. If you insist I'll do something, but I think I would prefer to leave it. I've removed the assert : pprice@tiger-sumire:~/LSST/external/galsim (tickets/DM-7114 %=) $ git sub-patch commit 26dc8bf89928ae3b68dd7fb8c23271941511a4c6 Author: Paul Price <price@astro.princeton.edu> Date: Wed Aug 3 12:36:24 2016 -0400   replace assertion with exception This is an exceptional condition that can occur sometimes (because it certainly does occur!), rather than something that should bring the world to a screaming halt. Reported upstream as PR #784: https://github.com/GalSim-developers/GalSim/pull/784   diff --git a/patches/DM-7114-assertion_to_exception.patch b/patches/DM-7114-assertion_to_exception.patch new file mode 100644 index 0000000..b64f1ff --- /dev/null +++ b/patches/DM-7114-assertion_to_exception.patch @@ -0,0 +1,15 @@ +diff --git a/src/hsm/PSFCorr.cpp b/src/hsm/PSFCorr.cpp +index a9b3260..21b92b1 100644 +--- a/src/hsm/PSFCorr.cpp ++++ b/src/hsm/PSFCorr.cpp +@@ -728,7 +728,9 @@ namespace hsm { + if (iy2 > ymax) iy2 = ymax; + dbg<<"y1,y2 = "<<y1<<','<<y2<<std::endl; + dbg<<"iy1,iy2 = "<<iy1<<','<<iy2<<std::endl; +- assert(iy1 <= iy2); ++ if (iy1 > iy2) { ++ throw HSMError("Bounds don't make sense"); ++ } + + // + /* Use these pointers to speed up referencing arrays */

John Parejko added a comment - 03/Aug/16 12:50 PM Ok, if string formatting is a pain, can you at least have the message say what the bad bounds are? I don't know what iy1,iy2 mean here, but something like: "Vertical data bounds (iy1,iy2) don't make sense." would be better, I think.

Paul Price added a comment - 03/Aug/16 3:45 PM Reworded the message, tested through Jenkins and merged to master.